Table 3.

Multilevel latent growth curve analyses¹²³

	Model 1		Model 2		Model 3
	Initial level of happiness	Change of happiness	Initial level of happiness	Change of happiness	Initial level of happiness	Change of happiness
PM2.5	−.01(.006)*	−.003(.004)
PM10			−.01(.003)**	−.001(.002)
NO2					−.01(.01)*	−.003(.004)
% of college graduates	.001(.002)	−.001(.001)	−.001(.002)	−.001(.001)	.002(.002)	−.001(.001)
Population densitya	–	–	-†	–	–	–
Mutualityb	−.09(.10)	–	−.19(.11)†	–	−.03(.10)	–
Male	.02(.03)	.04(.02)*	.02(.03)	−.04(.02)*	.02(.03)	−.04(.02)*
Family incomea	–	–	–		–
Non-intact familyb	−.06(.04)	–	−.06(.04)	–	−.06(.04)	–
Wave 1 Class rank	.02(.01)	.02(.01)	.02(.01)	.02(.01)	.02(.01)	.02(.01)
Wave 1 Negative life event	−.02(.008)*	.002(.004)	−.02(.008)*	.002(.004)	−.02(.008)*	.002(.004)
Wave 1 Low family cohesion	−.25(.03)**	.09(.02)**	−.25(.03)**	.09(.02)**	−.25(.03)**	.09(.02)**
Wave 1 Self-esteem	.04(.004)**	−.01(.003)**	.04(.004)**	−.01(.003)**	.04(.004)**	−.01(.003)**
Wave 1 Negative emotion	−.25(.02)**	.07(.01)**	−.25(.02)**	.07(.01)**	−.25(.02)**	.07(.01)**
Wave 1 Low peer support	−.17(.03)**	.03(.02)†	−.17(.03)**	.03(.02)†	−.17(.03)**	.03(.02)†
Wave 1 deviance	.01(.01)	−.001(.009)	.01(.01)	−.001(.009)	.01(.01)	−.001(.009)

n₁ = 2538; n₂ = 40

†p < .1;*p < .05; **p < .01

¹The presented number is the estimated coefficient and standard error is in the parentheses

²Model fit for each of the four models is acceptable, which may be obtained from the first author upon the request

³The basic growth curve model had marginal fit (χ² (1) = 13; CFI = .98; RMSEA = .068) with both initial level and change of happiness are all significant (initial level = 2.98**; change = −.10**).When estimated in three, both initial level and change of happiness remained significant with identical signs (initial level = 2.97**; change = −.067**)

^aThe effects of these two variables were estimated but very small because of original measurement scale. Hence, they were not presented here

^bThe effects of these two variables on the change of happiness are not estimated because these effects caused model estimation problems due to very small effect